aisingapore/gemma-sea-lion-v4-27b-it-gguf overview

{
  "metadata": {},
  "card_data": {
    "library_name": "transformers",
    "pipeline_tag": "text-generation",
    "base_model": [
      "aisingapore/Gemma-SEA-LION-v4-27B-IT"
    ],
    "language": [
      "en",
      "zh",
      "vi",
      "id",
      "th",
      "fil",
      "ta",
      "ms",
      "km",
      "lo",
      "my",
      "jv",
      "su"
    ],
    "license": "gemma",
    "base_model_relation": "quantized",
    "frontmatter": {
      "library_name": "transformers",
      "pipeline_tag": "text-generation",
      "base_model": [
        "aisingapore/Gemma-SEA-LION-v4-27B-IT"
      ],
      "language": [
        "en",
        "zh",
        "vi",
        "id",
        "th",
        "fil",
        "ta",
        "ms",
        "km",
        "lo",
        "my",
        "jv",
        "su"
      ],
      "license": "gemma",
      "base_model_relation": "quantized"
    },
    "hero_image_url": "v4-banner-gguf.png \"v4-banner-gguf\"",
    "summary": "Last updated: 2025-08-25 **SEA-LION** is a collection of Large Language Models (LLMs) which have been pretrained and instruct-tuned for the Southeast Asia (SEA) region. As of 25 Aug 2025, Gemma-SEA-LION-v4-27B-IT excels at Southeast Asian (SEA) tasks when compared to other open models with fewer than 200 billion parameters and demonstrates performance comparable to that of larger and top closed models. Gemma-SEA-LION-v4-27B-IT was quantized to create Gemma-SEA-LION-v4-27B-IT-GGUF. Gemma-SEA-LION-v4-27B-IT-GGUF can run on a laptop with Ollama with 16GB of memory. Gemma-SEA-LION-v4-27B-IT-GGUF inherits Gemma 3’s: ### Model Description  SEA-LION stands for *Southeast Asian Languages In One Network*. Quantization was performed on Gemma-SEA-LION-v4-27B-IT to produce optimized variants that reduce memory requirements while maintaining model quality. These quantized models support inference on a range of consumer-grade GPUs and are compatible with various inference engines. For tokenization, the model employs the default tokenizer used in Gemma 3 27B Instruct. This repo contains GGUF format models files for aisingapore/Gemma-SEA-LION-v4-27B-IT Model Weights included in this repository: > Take note that some GGUFs are split into parts. Most tools such as llama.cpp and those built on it do support split GGUFs, > pointing the platform to the first split will be sufficient for it to function. In the event where a merge is necessary, > it can be done using llama.cpp's gguf-split: ./gguf-split --merge ./path/to/first-split ./path/to/output-gguf More details: > gguf-split guide & README",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlibrary_name: transformers\npipeline_tag: text-generation\nbase_model:\n- aisingapore/Gemma-SEA-LION-v4-27B-IT\nlanguage:\n- en\n- zh\n- vi\n- id\n- th\n- fil\n- ta\n- ms\n- km\n- lo\n- my\n- jv\n- su\nlicense: gemma\nbase_model_relation: quantized\n\n---\n\n![Banner!](v4-banner-gguf.png \"v4-banner-gguf\")\n\n# Model Card for Gemma-SEA-LION-v4-27B-IT-GGUF\n\n<!-- Provide a quick summary of what the model is/does. -->\n\nLast updated: 2025-08-25\n\n**SEA-LION** is a collection of Large Language Models (LLMs) which have been pretrained and instruct-tuned \nfor the Southeast Asia (SEA) region.\n\nAs of 25 Aug 2025, Gemma-SEA-LION-v4-27B-IT excels at Southeast Asian (SEA) tasks when compared to other open models\n with fewer than 200 billion parameters and demonstrates performance comparable to that of larger and top closed models. \n Gemma-SEA-LION-v4-27B-IT was quantized to create Gemma-SEA-LION-v4-27B-IT-GGUF. \n Gemma-SEA-LION-v4-27B-IT-GGUF can run on a laptop with Ollama with 16GB of memory. \n\nGemma-SEA-LION-v4-27B-IT-GGUF inherits Gemma 3’s: \n\n- Large 128K context length \n\n- Image and text understanding capabilities, including document comprehension, visual Q&A, and image-grounded reasoning\n\n- Advanced function calling and structured outputs to allow for seamless integration into larger systems\n\n\n### Model Description\n\n<!-- Provide a longer summary of what this model is. -->\n\nSEA-LION stands for *Southeast Asian Languages In One Network*. \n\nQuantization was performed on Gemma-SEA-LION-v4-27B-IT to produce optimized variants that reduce memory requirements \nwhile maintaining model quality. These quantized models support inference on a range of consumer-grade GPUs \nand are compatible with various inference engines.\n\n\nFor tokenization, the model employs the default tokenizer used in Gemma 3 27B Instruct. \n\n\n- **Developed by:** AI Products Pillar, AI Singapore\n- **Funded by:** Singapore NRF\n- **Model type:** Decoder\n- **Context length:** 128k tokens\n- **Language(s):**  Burmese, English, Indonesian, Khmer, Lao, Malay, Mandarin, Tagalog, Tamil, Thai and Vietnamese\n- **License:** [Gemma Terms of Use](https://ai.google.dev/gemma/terms)\n- **Quantized from model:** Gemma-SEA-LION-v4-27B-IT\n\nThis repo contains GGUF format models files for aisingapore/Gemma-SEA-LION-v4-27B-IT\n\nModel Weights included in this repository:\n- [Gemma-SEA-LION-v4-27B-IT-Q4_K_M](https://huggingface.co/aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF/blob/main/Gemma-SEA-LION-v4-27B-IT-Q4_K_M.gguf) \n- [Gemma-SEA-LION-v4-27B-IT-Q8_0](https://huggingface.co/aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF/blob/main/Gemma-SEA-LION-v4-27B-IT-Q8_0.gguf)\n- [Gemma-SEA-LION-v4-27B-IT-BF16](https://huggingface.co/aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF/blob/main/Gemma-SEA-LION-v4-27B-IT-BF16-00001-of-00003.gguf)\n- [Gemma-SEA-LION-v4-27B-IT-mmproj](https://huggingface.co/aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF/blob/main/mmproj-Gemma-SEA-LION-v4-27B-IT.gguf)\n\n> Take note that some GGUFs are split into parts. Most tools such as llama.cpp and those built on it do support split GGUFs, \n> pointing the platform to the first split will be sufficient for it to function. In the event where a merge is necessary, \n> it can be done using llama.cpp's gguf-split: ./gguf-split --merge ./path/to/first-split ./path/to/output-gguf More details: \n> gguf-split guide & [README](https://github.com/ggerganov/llama.cpp/tree/master/examples/gguf-split)\n\n## How to Get Started with the Model\n\nUse the code below to get started with the model using **llama.cpp**\n\n**llama.cpp** (text-only)\n```python\n./llama-cli -hf aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF -p \"Write a poem about Dim Sum in Indonesian.\"\n```\n\n**llama.cpp** (image input)\n```python\nwget https://github.com/bebechien/gemma/blob/main/surprise.png?raw=true -O ~/Downloads/surprise.png\n./llama-gemma3-cli -hf aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF -p \"Describe this image in Malay.\" --image ~/Downloads/surprise.png\n\n```\n\n## Training Details\n\n**Training Data**\n\nThe dataset comprises Burmese, English, Indonesian, Khmer, Lao, Malay, Mandarin, Tagalog, Tamil, Thai and Vietnamese languages, collected from a mixture of sources including web data, code, open-source datasets, and synthetically generated datasets, amounting to a total of 500 billion tokens sampled from our bucket of 1 trillion tokens.\n\n\n## Evaluation\n\n<!-- This section describes the evaluation protocols and provides the results. -->\n\n### Performance Test Results\n\n| Quantized Variant | Model Size (GB) | VRAM Required (GB) | Time to First Token (s) | Tokens per Second |\n|-------------------|-----------------|--------------------|------------------------|-------------------|\n| BF16              | 54.0            | 71.1               | 2.03                   | 41.5              |\n| Q8_0              | 28.7            | 47                 | 3.179                  | 37                |\n| Q4_K_M            | 16.5            | 35.5               | 2.645                  | 59.9              |\n\nAdditional Remarks: \n\n- TTFT and Tokens per Second: measured with vllm on localhost and concurrency = 1.\n\n- GGUF served using llama.cpp with the following settings:\n\n  - Offload all layers to GPU, Context Length 128K\n\n- Reported results are the median (p50) values, calculated across 10 requests.\n\n- Input size 4K, output 1K\n\n- Tests conducted on a system with an NVIDIA H100 GPU\n\n\n### Out-of-Scope Use\n\n<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->\n\nThe model has not been aligned for safety. Developers and users should perform their own safety \nfine-tuning and related security measures. In no event shall the authors be held liable for any claims, damages, or other liabilities arising from the use of the released weights and codes.\n\n\n## Bias, Risks, and Limitations\n\n<!-- This section is meant to convey both technical and sociotechnical limitations. -->\n\n*The model was not tested for robustness against adversarial prompting.* It is important for users to be aware that our model exhibits certain limitations that warrant consideration. \nLike many LLMs, the model can hallucinate and occasionally generates irrelevant content, \nintroducing fictional elements that are not grounded in the provided context. \nUsers should also exercise caution in interpreting and validating the model's responses \ndue to the potential inconsistencies.\n\n**Limitations**\n\nIn terms of vision capability, Gemma-SEA-LION-v4-27B-IT has been trained and fine-tuned exclusively on the text back-end.\nAs a result, its vision capabilities are expected to be comparable to those of Gemma 3 IT 27B, \nand may not exhibit significant improvements or differences in this area. [🤗 google/gemma-3-27b-it](https://huggingface.co/google/gemma-3-27b-it )\n\n\n\n## More Information\n\nThis is the repository for the commercial instruction-tuned model. \nThe model has not been aligned for safety. Developers and users should perform their own safety \nfine-tuning and related security measures. In no event shall the authors be held liable \nfor any claims, damages, or other liabilities arising from the use of the released weights and codes.\n\nAI Singapore is a national programme supported by the National Research Foundation, Singapore and hosted by the National University of Singapore. \nAny opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of the National Research Foundation or the National University of Singapore.\n\nFor more info, please contact us at sealion@aisingapore.org\n\n\n## Team\n\nAntonyrex Sajeban, Chan Hok Teng Adwin, Cheng Zi Yi Nicholas, Choa Hsueh Mei Esther, Heng Jonathan, Huang Yuli, Hulagadri Adithya Venkatadri, Jann Railey Estrada Montalan, \nKang Siow Wei  Bryan, Lau Wayne, Lee Chwan Ren, Leong Wai Yi, Leong Wei Qi, Limkonchotiwat Peerat, Muhammad Ridzuan Bin Mokhtar, Nagarajan Karthik, Ng Boon Cheong  Raymond,\nNgee Chia Tai, Ngui Jian Gang, Nguyen Thanh Ngan, Ong Jin Jie Brandon, Ong Tat-Wee David, Ong Zhi Hao, Pereira Mark, Rengarajan Hamsawardhini, Susanto Yosephine, \nSutaveephamochanon Anocha, Tan Choon Meng, Tan Chor Phin Evelyn, Tan Siao Wei Jessica, Teng Kok Wai Walter, Teo Eng Sipp Leslie, Tjhi William, Yeo Yeow Tong, Yong Xianbin, \nLiew Rachel, Liu Bing Jie Darius, Teo Wei Yi, Zhou Lin (NCS), Gopalakrishnan Roshan (NCS), Anda Cuahtemoc (NCS), Sri Devi Wijaya (NCS),  Nandi Partha (NCS), \nElliott Chris (Google), Mohseni Mohammadreza (Google), Sharan Mayank (Google), Wei Fanny (Google), Tang Jiuqiang (Google), Xu Xiang (Google), Yu Ting (Google), \nLoh Michelle (Google), Mangal Saurabh (Google), Mukherjee Pratyusha (Google), Sim Stephanie (Google)\n\n\n## Acknowledgement\n\nThis project is supported by the National Research Foundation Singapore and Infocomm Media Development Authority (IMDA), \nSingapore under its National Large Language Model Funding Initiative.\n\n## Contact\n\nsealion@aisingapore.org",
    "related_quantizations": []
  },
  "tags": [
    "transformers",
    "gguf",
    "text-generation",
    "en",
    "zh",
    "vi",
    "id",
    "th",
    "fil",
    "ta",
    "ms",
    "km",
    "lo",
    "my",
    "jv",
    "su",
    "base_model:aisingapore/Gemma-SEA-LION-v4-27B-IT",
    "base_model:quantized:aisingapore/Gemma-SEA-LION-v4-27B-IT",
    "license:gemma",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 2,
  "downloads": 152,
  "gated": false,
  "private": false,
  "last_modified": "2025-12-02T02:42:17.000Z",
  "created_at": "2025-08-19T02:40:44.000Z",
  "pipeline_tag": "text-generation",
  "library_name": "transformers"
}

{
  "_id": "68a3e42c5ace8d96aa9cd74d",
  "id": "aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF",
  "modelId": "aisingapore/Gemma-SEA-LION-v4-27B-IT-GGUF",
  "sha": "74d2fb09972765d91ac6142154861079ae3256f4",
  "createdAt": "2025-08-19T02:40:44.000Z",
  "lastModified": "2025-12-02T02:42:17.000Z",
  "author": "aisingapore",
  "downloads": 152,
  "likes": 2,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "transformers",
  "siblings_count": 9
}